Such a microscope design comes under the category known as SPIM microscopes (SPIM—Selective Plane Illumination Microscopy). In contrast to confocal laser scanning microscopy (LSM), in which a three-dimensional sample is scanned point by point in individual planes at different depths and the image information obtained thereby is subsequently assembled to form a three-dimensional image of the sample, the SPIM technology is based on wide-field microscopy and permits the imaging of the sample on the basis of optical sections through various planes of the sample.
The advantages of the SPIM technology consist, among others, in the greater speed at which the image information is detected, the reduced risk of bleaching of biological samples, and a greater penetration depth of the focus into the sample.
The principle of SPIM technology is that fluorophores contained in the sample originally or added to it for contrasting are excited with laser light, the laser radiation being shaped into a so-called light sheet. The light sheet is used to illuminate a selected plane in the depth of the sample in the sample region, and an imaging lens system is used to obtain an image of this sample plane in the form of an optical section.
First modern approaches to SPIM technology are described by A. H. Voie et al., Journal of Microscopy, Vol. 170 (3), pp. 229-236, 1993. Here, the fundamentals of modern SPIM technology are explained, in which a coherent light source is used to illuminate a sample, the light sheet being produced with the aid of a cylindrical lens. Arranged normal to the propagation direction of the light sheet, which has a finite thickness, though, are detection means comprising an imaging lens system and a camera.
In recent years, the technology was developed further, especially with regard to its application in fluorescence microscopy. For example, DE 102 57 423 A1 and, based on it, WO2004/053558A1 describe methods in which a light-sheet-like illumination is produced due to a relative movement between a line-shaped field of light and the sample to be examined. The light-sheet-like illumination is formed by the field of light being repeated in a temporal succession so as to be lined up side by side due to the relative movement. In this way, though, shadows are formed within the sample plane to be examined, on account of parts of the sample that lie in the direction of illumination and are not transparent to the illuminating light. Similar setups are also described by Stelzer et al., Science (305) pp. 1007-1009 (2004), and Reynaud et al., HFSP Journal 2, pp. 266 (2008).
Instead of a purely static light sheet, for the generation of which a cylindrical lens system is used, it is possible to produce a quasi-static light sheet by rapidly scanning the sample with a rotationally symmetric light beam. The integration time of the camera on which the sample is imaged is chosen so that the scan is completed within the integration time. Such setups are described, e.g., by Keller et al., Science (322), pp. 1765 (2008), and Keller et al., Current Opinion in Neurobiology 18, pp. 1-9 (2009).
All the setups and methods known in prior art, however, have more or less grave disadvantages, which restrict the use of the SPIM technology in the commercial sphere, where it is important, among other things, to achieve a high user friendliness of the microscopes and, as a rule, a high throughput, with a great number of samples having to be examined within a relatively short time. Essential disadvantages are described below.
In most of the setups using SPIM technology that have been implemented so far, e.g., those according to DE 102 57 423 A1 and WO2004/053558A1, the mere variation of the image field size for detection—e.g., switching from an image field size providing a good overview of the sample to a detail region—is rather a complex and time-consuming affair. It can only be implemented by a change of the detection objective. This affects the sample space unfavorably, which may have a particularly negative effect in case of a horizontal detection beam path. In the worst case, it also involves the removal and emptying of the sample chamber. After this, refocusing is necessary as a rule. Moreover, the sample is unnecessarily heated or cooled.
An improvement is described by Becker et al., Journal of Biophotonics 1 (1), pp. 36-42 (2008). Here, the detection beam path is arranged vertically, so that a change of the image field size can be carried out without any substantial interaction with the sample chamber volume. The detection objective can be put into the sample chamber and taken out from above in a simple manner. Nevertheless, slight interactions with the sample chamber and, thus, indirectly with the sample cannot be avoided.
Adaptation of the image field size is even simpler if zoom detection objectives are used. Such a setup is described, e.g., by Santi et al., Biotechnics 46, pp. 287-294 (2009). Here, a commercial microscope, the Olympus MVX10, which has a zoom objective, is used for detection. This, too, is inserted into the sample chamber from above, which is, as a rule, filled with an immersion liquid, so that, here again, there are slight interactions with the sample chamber when the zoom function of the objective is working or when the focus is adjusted, alone because of the motorized shifting of the lenses, which may cause vibrations that may transmit to the liquid in the sample chamber.
If the image field size for detection is changed, it is also desirable to adapt the illumination-side image field, i.e. to adapt the extension of the light sheet along the transverse axis Y and the detection axis Z. In prior art, this adaptation has so far been implemented by the use of interchangeable diaphragms and/or beam expanders, as described, e.g., by Keller et al., Science 322, pp. 176 ff. (2008), and by Huisken et al., Optics Letters 32 (17), pp. 2608-2610 (2007). The flare occurring in case of diaphragms causes light losses, whereas the use of beam expanders reduces flexibility, since exchanging them is rather laborious.
While in the classical way, as described, e.g., in WO2004/053558A1, the light sheet is produced via cylindrical lenses arranged in the beam path, the recent state of prior art, as described, e.g., in the above-mentioned article by Keller et al., Science 322, pp. 176 ff. (2008), uses setups in which no static light sheet is produced but merely a quasi-static light sheet, where the sample is rapidly scanned by a rotationally symmetric light beam. ‘Rapidly’ means that the integration time of the spatially resolving array detector used as a rule, e.g., a camera with CCD chip or CMOS chip, is chosen so that the light beam scans the sample region corresponding to the quasi-static light sheet within this integration time. The integration time—which, in the camera, e.g., corresponds to the shutter opening time—and the scanning frequency or scanning time of the light beam may, as a rule, be set independently of each other, so that the scanning time can be adapted to a fixed integration time. As scanning with a rotationally symmetric light beam also produces a light sheet, at least in effect, this approach is also subsumed under the generation of a light sheet.
Both kinds of light sheet generation have advantages and disadvantages. With the use of cylindrical lenses, e.g., there is less of a load on the sample, because the intensity with which the sample is irradiated can be selected at a lower level while nevertheless the same dose is achieved as in case of scanning. Also, the use of cylindrical lenses is well suitable for recording image sequences in fast succession within very short times, since the speed is not limited by movable elements in the illumination beam path. In particular, a stroboscope-like illumination can be implemented very well with the use of cylindrical lenses. In scanning, the swiveling scanning mirror used, as a rule, is apt to be the speed-limiting element. If plain scanning is combined with angular scanning, i.e. illumination from different angles, in order to reduce banding as described, e.g., in DE 10 2007 015 063 A1, there is a risk that beat artefacts will be produced if the scanners for light sheet angle scanning and position scanning are not matched, i.e. not synchronized.
Advantages of light sheet generation by scanning are given by, among other things, the fact that it permits a more homogeneous illumination of the sample, so that quantitative image evaluations are possible as well, which by the use of a cylindrical lens system can be achieved only approximately by flaring through a diaphragm, which entails light losses. Moreover, a flexible choice of the maximum deflection of the scanner will permit the size of the image to be adapted with high flexibility. Scanning reduces the spatial coherence of the excitation light, which also leads to a reduction of banding. Finally it is possible, by special modulations of the light source, e.g. with an AOTF, to project grid patterns into the sample.
In other setups described in prior art, the sample is illuminated from both sides, from opposite directions along the illumination axis X. In the setup described by Santi et al., Biotechnics 46, pp. 287-294 (2009), the sample is illuminated simultaneously from both sides. For many kinds of samples, such as embryos of the fruit fly (Drosophila), such a setup is not of advantage, because in this way scattering and non-scattering image portions are combined in an unfavorable way. Huisken et al., Optics Letters 32 (17), pp. 2608-2610 (2007), and Becker et al., Journal of Biophotonics 1 (1), pp. 36-42 (2008), describe setups that illuminate the sample sequentially, i.e. alternately from the two directions along the illumination axis X, which is more favorable for the sample mentioned above. For switching back and forth between the two illumination directions, a vibration-producing shutter or a rotating mirror is used, so that the times required for switching are relatively long.
Keller et al., in Science 322, pp. 1765 ff. (2008) and in Current Opinion in Neurobiology, 18, pp. 1-9 (2009), describe an SPIM setup in which the illumination and/or detection objective is mounted on a piezo motor, which permits focusing. Here, then, setting the focusing distance is accomplished via a displacement of the entire objective. In particular, the distance of the front lens from the image plane is not maintained, so that an interaction with the sample chamber is possible. This applies especially to horizontal detection beam paths with immersed detection objectives: Here, the necessary movement of the objective entails tightness problems. On the other hand, a movable element in the sample space is disturbing in general, as the user may need space there for diverse means for feeding to the sample chamber. The vibrations occurring during the movement of the objective may be unfavorably transmitted to the sample, since the space between the objective and the sample is occupied by a liquid rather than by air.
If scanners are used for producing the light sheet, the imaging of the scanner into the pupil of the illumination objective is not optimal, as a rule, so that the plain position scanning is superposed by portions of angular scanning.
Also known in prior art are setups in which the detection beam path is split up into two branch beam paths; this is described, e.g., in the two publications by Keller et al. mentioned above. For the beam splitting one uses beam splitters which transmit part of the light into one branch beam path and reflect the other part of the light into the other branch beam path. For this purpose one uses common dichroic filters having a relatively small thickness of less than 2 mm, which are arranged in a divergent part of the detection beam path. The advantage of such an arrangement is that in the direction of transmission there occur hardly any artefacts caused by astigmatism. In the direction of the reflected light, however, image artefacts such as astigmatism or also defocusing do occur, due to surface tensions at the dichroic filter, which can be caused, e.g., by the coating, or by improper installation. Another way of splitting into two branch detection beam paths is described by Huisken et al. in Optics Letters 32, pp. 2608-2610 (2007). Here, the dichroic filter is located at infinity (related to the beam path), so that, here again, the problems occurring in transmission are minimized. As far as the reflected branch beam path is concerned, though, the problem of surface tensions may occur here, too, if conventional dichroic filters are used.